The present invention relates to multiple sensor integration (MSI), multiple sensor tracking (MST), and more particularly to the association of tracks generated by three collocated dissimilar sensor systems: radar, infrared (IR), and electronic support measures (ESM).
Radar Systems are commonly used to detect and track targets. Such a radar system includes a radar sensor and a computer. The radar sensor radiates electromagnetic energy, and then detects and measures the echoes returned from reflecting objects. The radar computer processes the radar measurements and uses them to maintain a collection of radar "tracks", where each such track estimates the position, velocity and other attributes of a target, enabling the prediction of the target's position. A radar system typically provides target range, azimuth, range rate, and azimuth rate information for each radar track. Some radars have height finding capability, in which case their radar tracks also include elevation information.
In many military applications, such as ship self-defense systems, other types of sensors, including IR sensors and ESM sensors, may also be used in addition to radar to enhance the detection and tracking of targets and to support the designation of targets to weapon systems.
An infrared sensor system detects infrared radiation. Such a sensor system uses radiation emitted by targets to detect and track targets. An IR track includes azimuth and elevation data, but no range data.
An ESM sensor system detects radio frequency energy. Such a sensor system includes a radio frequency receiver detecting and monitoring electromagnetic emissions from targets, and a processor estimating the waveform parameters, the angle-of-arrival, and the identification of the received signals. An ESM track typically provides bearing (angle-of-arrival), identification and waveform information, but does not include range or elevation information.
The integration of collocated multiple sensors is needed in ownship point defense applications for many reasons. The ownship multi-sensor integration enhances the target acquisition and tracking performance. A target undetectable by the ownship radar due to its small cross section or its position in the radar's multipath nulls may be detected and tracked by the ownship IR sensor, or, in the presence of radio frequency emissions from the target, by the ownship ESM sensor. A track undetectable by the ownship IR sensor due to unfavorable atmospheric conditions may be detected and tracked by the ownship radar and ESM sensor. A track undetectable by the ownship ESM sensor due to the lack of radio frequency emissions may be detected and tracked by the ownship radar and IR sensor. The ownship radar estimate of a target's position and identification may be significantly improved when that target is tracked by all sensors and the multi-sensor integration is performed. The improvement in that case is a result of using the more accurate target azimuth and elevation included in the IR track, and the more accurate target identification included in its ESM track.
The ownship multi-sensor integration may also be dictated by the need to support new or multiple weapon systems. Successful designation of a target to a weapon system whose main guidance is based on radio frequency emissions may only be performed if that target is known to be actively emitting.
When multiple targets are tracked by multiple ownship sensors, the number of tracks generated by each target could be as high as the number of sensors, and thus in the presence of numerous targets it becomes imperative to generate and maintain a coherent tactical picture of the environment, in which each target is represented by a unique global track. In a dense environment, an accurate tactical picture free of redundancies, conflicts and ambiguities can only be obtained by employing an automatic data fusion algorithm (procedure), which correctly and efficiently combines data generated by the ownship sensors into a global (central) track file. When individual sensor track files are maintained for each sensor, the multi-sensor integration typically involves a track-to-track association process which provides links (associations) between the sensor tracks potentially representing the same targets, and then fuses the data of the associated tracks.
Any multi-sensor track-to-track association employs an assignment scheme between the members of the different track files that it operates on. As an example, a radar/ESM track association operates on a radar track file and an ESM track file, and may assign a radar track to each ESM track, if certain association criteria are met. Such an assignment scheme, operating on two track files, is 2-dimensional. To determine the associations between a set of radar tracks and a set of ESM tracks means to establish a set of rules assigning the radar tracks to ESM tracks or the ESM tracks to radar tracks. The assignments could be one-to-one (when each radar track may only be linked to at most one ESM track and each ESM track may only be linked to at most one radar track) or many-to-one (e.g., when a radar track may be linked to more than one ESM track).
In case of a radar/IR/ESM track association, its assignment scheme operates on three track files: the radar track file, the IR track file and the ESM track file. Such an assignment scheme, operating on three track files, is 3-dimensional. Ideally, for each target which generates a track in each one of the three track files, those tracks should be associated to each other. When a tri-sensor track association procedure is based on first making 2-dimensional assignments (i.e., pairwise associations), conflicts and ambiguities may arise when the final 3-dimensional assignments (which we call global associations) are made. To illustrate such a conflict, an IR track and an ESM track may have pairwise links to the same radar track but not to each other. A new 3-dimensional assignment method which resolves such conflicts and ambiguities is the subject of this invention.
The following references are related to the subject matter of this invention and are described below:
[1] J. W. Thomas, A Radar-ESM Correlation Discriminant Using Bearing History, JHU/APL Report F3A-79-2-085, March 1979. PA1 [2] G. V. Trunk and J. D. Wilson, Association of DF Bearing Measurements with Radar Tracks, IEEE Transactions on Aerospace and Electronic Systems, Vol. AES-23, No. 4, July 1987, pp. 438-447. PA1 [3] C. L. Morefield and C. M. Peterson, Data Association Algorithms for Large Area Surveillance, ORSA/TIMS Meeting, ORINCON Corp., May, 1978, La Jolla, Calif. 1978, DTIC AD-A086606/1. PA1 [4] C. B. Chang and L. C. Youens, Measurement Correlation For Multiple Sensor Tracking in a Dense Target Environment, Technical Report 549, Lincoln Laboratory, MIT, Lexington, Mass., Jan. 20, 1981, DTIC AD-A098001. PA1 [5] C. L. Bowman and M. Gross, Multi-Sensor Multi-Platform Track Association Algorithm Using Kinematics and Attributes, IEEE National Aerospace and Electronics Conference, Dayton, May 20-24, 1985, vol. 1, pp. 204-208. PA1 [6] S. S. Blackman, Multiple-Target Tracking with Radar Applications, Artech House, Norwood, Mass., 1986, pp. 363-367. PA1 [7] E. Davis, The Mission Avionics Sensor Synergism (MASS) Program, Proceedings of the 1987 Tri-Service Data Fusion Symposium, 9-11 June 1987, JHU/APL, Laurel, MD, Published by NADC, Warminster, Pa. 1987, pp. 366-372. PA1 [8] W. R. Ditzler, D. W. Cowan, A. G. Sutton and E. Benites, Integrated Multisensor Tracking, Proceedings of the U.S. DoD Tri-Service Combat Identification Systems Conference, Arlington, Va. 1986. PA1 [9] E. Benites, P. R. Decker, W. R. Ditzler and A. G. Sutton, A Demonstration of Multisensor Tracking, Proceedings of the 1987 Tri-Service Data Fusion Symposium, 9-11 June 1987, JHU/APL, Laurel, MD, Published by NADC, Warminster, Pa., 1987, pp. 303-311.
Radar/ESM track association algorithms have been known for some time [1] and continue to be the subject of research [2]. Algorithms are known which allow multiple ESM tracks to link with a single radar track but not vice versa, keep positional association (PA) lists of candidate radar tracks for association with each ESM track based on historical azimuth proximity, form links only when high confidence associations are obtained, and resolve ambiguities based on a sequential probability ratio test. The algorithm described in [1] keeps a running bearing difference history for each candidate track pair, and uses it, as a fine discriminant, to resolve association ambiguities, allowing one-to-one associations only. The logic proposed in [2] allows multi-links and uses a multiple hypothesis testing method based on the chi-squared distributed observed squared error in bearing.
One association algorithm, designed to integrate a radar system and an ESM system in support of a missile system homing on target radio frequency emissions, uses PA lists, allows multi-links reflecting different sensor resolution capabilities, employs a running time average absolute value bearing difference history for fine discrimination, and makes immediate association decisions for all ESM tracks when they receive updates, and periodically each radar scan.
References [3-9] describe multi-sensor track association algorithms which could integrate three or more sensors. The logic proposed in [6], and the logics [5] and [7-9], are based on sequentially pairwise associations involving 2-dimensional assignment logics, and fusion of tracks into central level tracks. Such logics may not detect global conflicts or ambiguities and are likely to propagate association errors. The algorithm described in [8] and [9] is a modification of [7] which employs a one-to-one assignment logic, a clustering and centroiding logic, and a splitting and merging logic to account for the different resolution capabilities of the radar and IR sensor. The computational complexity of such an algorithm could be prohibitive in applications using limited data processing capabilities. Reference [3] shows that the association problem is an assignment problem, notes that no efficient method for finding the optimal solution to the multi-dimensional (i.e., of dimension at least 3) assignment problem is known, and proposes a suboptimal solution recommended for low density environments, and allowing multi-links. The multi-link structure thus obtained may not be consistent with the specific target resolution capabilities of the different sensors. The association algorithm proposed in [4] is based on using Kalman filter residuals as measures of track similarity and reducing the one-to-one multi-dimensional assignment problem to a 0-1 integer programming problem. The run time of the integer program is shown, however, to grow exponentially with the number of potential correlations.